> ## Documentation Index
> Fetch the complete documentation index at: https://agentset-feat-agent-first-api-mcp.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Search

> Search your documents with Agentset

Search your namespace to retrieve relevant chunks from your documents. Agentset uses semantic and keyword search to return the most relevant results.

## Basic search

Pass a query string to search your namespace and retrieve chunks.

<CodeGroup>
  ```typescript TypeScript theme={null}
  import { Agentset } from "agentset";

  const agentset = new Agentset({
    apiKey: process.env.AGENTSET_API_KEY,
  });

  const ns = agentset.namespace("YOUR_NAMESPACE_ID");

  const results = await ns.search("What is machine learning?");

  for (const result of results) {
    console.log(result.text);
  }
  ```

  ```python Python theme={null}
  import os
  from agentset import Agentset

  client = Agentset(
      namespace_id="YOUR_NAMESPACE_ID",
      token=os.environ["AGENTSET_API_KEY"],
  )

  results = client.search.execute(query="What is machine learning?")

  for result in results.data:
      print(result.text)
  ```
</CodeGroup>

## Limiting results

Control the number of results returned with `topK`. The default is 10, with a maximum of 100.

<CodeGroup>
  ```typescript TypeScript theme={null}
  const results = await ns.search("quarterly revenue", {
    topK: 20,
  });
  ```

  ```python Python theme={null}
  results = client.search.execute(
      query="quarterly revenue",
      top_k=20,
  )
  ```
</CodeGroup>

## Reranking

Agentset reranks results by default using a cross-encoder model to improve relevance. You can adjust reranking behavior or disable it.

<CodeGroup>
  ```typescript TypeScript theme={null}
  const results = await ns.search("product roadmap", {
    topK: 50,
    rerank: true,
    rerankLimit: 10,
  });
  ```

  ```python Python theme={null}
  results = client.search.execute(
      query="product roadmap",
      top_k=50,
      rerank=True,
      rerank_limit=10,
  )
  ```
</CodeGroup>

| Parameter     | Default                  | Description                                 |
| :------------ | :----------------------- | :------------------------------------------ |
| `rerank`      | `true`                   | Enable or disable reranking                 |
| `rerankLimit` | Same as `topK`           | Number of results to return after reranking |
| `rerankModel` | `cohere:rerank-v4.0-pro` | Model used for reranking                    |

### Available rerank models

| Model                             | Description                            |
| :-------------------------------- | :------------------------------------- |
| `cohere:rerank-v4.0-pro`          | Most capable Cohere reranker (default) |
| `cohere:rerank-v4.0-fast`         | Faster Cohere v4.0 reranker            |
| `cohere:rerank-v3.5`              | Cohere v3.5 reranker                   |
| `cohere:rerank-english-v3.0`      | English-optimized Cohere reranker      |
| `cohere:rerank-multilingual-v3.0` | Multilingual Cohere reranker           |
| `zeroentropy:zerank-2`            | Latest ZeroEntropy reranker            |
| `zeroentropy:zerank-1`            | ZeroEntropy reranker                   |
| `zeroentropy:zerank-1-small`      | Smaller, faster ZeroEntropy reranker   |

## Minimum score threshold

Filter out low-relevance results by setting a minimum score.

<CodeGroup>
  ```typescript TypeScript theme={null}
  const results = await ns.search("API documentation", {
    minScore: 0.7,
  });
  ```

  ```python Python theme={null}
  results = client.search.execute(
      query="API documentation",
      min_score=0.7,
  )
  ```
</CodeGroup>

## Search modes

Agentset supports two search modes.

<CodeGroup>
  ```typescript TypeScript theme={null}
  const results = await ns.search("error code 500", {
    mode: "keyword",
  });
  ```

  ```python Python theme={null}
  results = client.search.execute(
      query="error code 500",
      mode="keyword",
  )
  ```
</CodeGroup>

| Mode       | Description                                                    |
| :--------- | :------------------------------------------------------------- |
| `semantic` | Uses embeddings to find semantically similar content (default) |
| `keyword`  | Traditional keyword-based search                               |

## Response structure

Each result includes an ID, relevance score, and text content. Metadata is included by default.

```json theme={null}
{
  "success": true,
  "data": [
    {
      "id": "chunk_abc123",
      "score": 0.92,
      "text": "Machine learning is a subset of artificial intelligence...",
      "metadata": {
        "filename": "ml-guide.pdf",
        "filetype": "application/pdf",
        "file_directory": "/documents"
      }
    }
  ]
}
```

### Excluding metadata

<CodeGroup>
  ```typescript TypeScript theme={null}
  const results = await ns.search("user authentication", {
    includeMetadata: false,
  });
  ```

  ```python Python theme={null}
  results = client.search.execute(
      query="user authentication",
      include_metadata=False,
  )
  ```
</CodeGroup>

## Next steps

* [Simple RAG](/search-and-retrieval/simple-rag) — Generate answers grounded in search results
* [Agentic Search](/search-and-retrieval/agentic-search) — Let the model search on its own for complex questions
* [Filtering](/search-and-retrieval/filtering) — Narrow results by document metadata
* [Ranking](/search-and-retrieval/ranking) — Configure how results are scored